Three-dimensional Gaussian model to define brain metastasis limits on 11C-methionine PET

Since 11C-methionine (MET) heavily accumulates in brain tumors, PET with MET (MET-PET) is proposed for the image-guided planning of their targeted therapy. Determination of bulk tumor limits is therefore a crucial component of MET-PET image analysis. We aimed at validating a Gaussian model of tumor delineation on MET-PET. We choose MET-PET and MRI data obtained in brain metastases to adjust the model. Indeed, MRI limits of these non-infiltrative hypermetabolic brain lesions are efficiently used for their curative treatment.

We developed a three-dimensional (3D) Gaussian model that relates the tumor-limit-defining threshold to maximum and mean count values in the defined tumor volume and to mean count values in a reference region. To adjust the model to experimental data, we selected 25 brain metastases following these criteria: (i) no surgery or classical radiotherapy within 6 months, (ii) no previous radiosurgery, (iii) MET-PET and MRI acquired within a 48-h interval, (vi) necrosis representing less than 25% of tumor volume on MRI. We applied a progressive thresholding procedure on MET-PET so as to match tumor limits on contrast-enhanced co-registered MRI.

In 22 tumors, a match could be reached between tumor margins on MET-PET and MRI. The relation between mean, maximum and threshold values closely fits the 3D-Gaussian model function. We found a quadratic relation between the mean-to-threshold ratio and the maximum-to-cerebellum activity ratio.

A 3D-Gaussian model may describe the limits of MET uptake distribution within brain metastases, providing a simple method for metabolic tumor delineation.